High degree of genetic polymorphism in apolipoprotein(a) associated with plasma lipoprotein(a) levels in Japanese and Chinese populations

We have developed a sensitve, high-resolution method for the analysis of the apolipoprotein(a) [apo(a)] isoforms using sodium dodecyl sulfate (SDS)-agarose/ gradient polyacrylamide gel electrophoresis. In an analysis of the genetic polymorphism of apo(a) isoforms and their relationship with plasma lipoprotein(a) [Lp(a)] levels in Japanese and Chinese, this method identified 25 different apo(a) isoforms and detected one or two apo(a) isoforms in more than 99.5% of the individuals tested. The apparent molecular weights of the apo(a) isoforms ranged from 370 kDa to 950 kDa, and 22 of the 25 different apo(a) isoforns had a higher molecular weight than of apo B-100. Studies on Japanese families confirmed the autosomal codominant segregation of apo(a) isoforms and the existence of a null allele at the apo(a) locus. The observed frequency distribution of apo(a) isoform phenotypes fit the expectations of the Hardy-Weinberg equilibrium in both the Japanese and Chinese populations. Our data indicate the existence of at least 26 alleles, including a null allele, at the apo(a) locus. The frequency distribution patterns of the apo(a) isoform alleles in Japanese and Chinese were similar to each other and also similar to that of apo(a) gene sizes reported in Caucasian American individuals. The average heterozygosity at the apo(a) locus was 92% in Japanese and 93% in Chinese. A highly significant inverse correlation was observed between plasma Lp(a) levels and the size of apo(a) isoforms in both the Japanese (r=-0.677, P=0.0001) and the Chinese (r=-0.703, P=0.0001). A highly skewed distribution of Lp(a) concentrations towards lower levels in the Japanese population may be explained by high frequencies of alleles encoding large apo(a) isoforms and the null allele.     

Characterization of high density lipoprotein particles in familial apolipoprotein A-I deficiency

Our aim was to characterize HDL subspecies and fat-soluble vitamin levels in a kindred with familial apolipoprotein A-I (apoA-I) deficiency. Sequencing of the APOA1 gene revealed a nonsense mutation at codon -2, Q[-2]X, with two documented homozygotes, eight heterozygotes, and two normal subjects in the kindred. Homozygotes presented markedly decreased HDL cholesterol levels, undetectable plasma apoA-1, tuboeruptive and planar xanthomas, mild corneal arcus and opacification, and severe premature coronary artery disease. In both homozygotes, analysis of HDL particles by two-dimensional gel electrophoresis revealed undetectable apoA-I, decreased amounts of small alpha-3 migrating apoA-II particles, and only modestly decreased normal amounts of slow alpha migrating apoA-IV- and apoE-containing HDL, while in the eight heterozygotes, there was loss of large alpha-1 HDL particles. There were no significant decreases in plasma fat-soluble vitamin levels noted in either homozygotes or heterozygotes compared with normal control subjects. Our data indicate that isolated apoA-I deficiency results in marked HDL deficiency with very low apoA-II alpha-3 HDL particles, modest reductions in the separate and distinct plasma apoA-IV and apoE HDL particles, tuboeruptive xanthomas, premature coronary atherosclerosis, and no evidence of fat malabsorption.     

Characterization of apolipoprotein M isoforms in low-density lipoprotein

Apo M is a recently discovered human lipoprotein thought to be involved in the metabolism of lipids and lipoprotein particles. Here, a proteomic approach was applied to examine the glycosylation pattern of apo M in human LDL. We treated LDL proteins with N-glycosidase or neuraminidase, studied mobility shifts of Apo M by two-dimensional gel electrophoresis, and different isoforms were then identified with mass spectrometry. This way, we demonstrated the presence of five isoforms of apo M in LDL: three that are both N-glycosylated and sialylated, one that is N-glycosylated but not sialylated, and one that is neither N-glycosylated nor sialylated. As judged from the examination of LDL from 20 healthy human subjects, the three N-glycosylated and sialylated forms are most abundant (80-100% of the total apo M in LDL) whereas the unsialylated and unglycosylated variants constitute at most 20%. Comparative analysis showed that the same five isoforms of apo M are also present in HDL. Further studies aiming at elucidating the role of apo M in health and disease will have to take this polymorphism of apo M proteins into account.     

Isolation of apolipoprotein(a) from lipoprotein(a)

An easy method was developed for the rapid and selective isolation of apo(a) from human plasma Lp(a). This procedure was applied to a "low density" Lp(a) subspecies (usually found in the density interval 1.050 to 1.070 g/ml) from a single individual whose apo(a) was of a size smaller than apoB-100. After reduction with 0.01 M dithiothreitol, apo(a) was separated from the Lp(a) particle by rate zonal centrifugation on a 7.5-30% NaBr density gradient. Two completely water-soluble products were recovered: apo(a), which contained less than 1% each of phospholipid and cholesterol, remained at the bottom of the gradient, and a lipid-rich floating LDL-like particle which contained apoB but not apo(a) and which we referred to as Lp(a-). The separation of these two components was also achieved by subjecting reduced Lp(a) to electrophoresis on 2.5-16% polyacrylamide gradient gels. However, dissociation of reduced Lp(a) could not be achieved by gel filtration in either low or high salt solutions. These observations indicate that apo(a) is associated to Lp(a) by non-covalent interactions in addition to its disulfide linkage to apoB. The latter is sensitive to chemical reduction whereas the former are broken through the action of a gravitational or electrical field.     

High levels of human apolipoprotein A-I in transgenic mice result in increased plasma levels of small high density lipoprotein (HDL) particles comparable to human HDL3

Five lines of transgenic mice, which had integrated the human apolipoprotein (apo) A-I gene and various amounts of flanking sequences, were established. Normally, apoA-I is expressed mainly in liver and intestine, but all of the transgenic lines only expressed apoA-I mRNA in liver, strongly suggesting that 256 base pairs of 5'-flanking sequence was sufficient for liver apoA-I gene expression but that 5.5 kilobase pairs was not sufficient for intestinal expression. Mean plasma levels of human apoA-I varied in different lines from approximately 0.1 to 200% of normal mouse levels. This was not dependent on the amount of flanking sequence. Lipoprotein levels were studied in detail in one of the lines with a significantly increased apoA-I pool size. In one study, the total plasma apoA-I level (mouse plus human) was 381 +/- 43 mg/dl in six animals from this line, compared to 153 +/- 17 mg/dl in matched controls. Total and high density lipoprotein cholesterol (HDL-C) levels were increased 60% in transgenic animals, compared to controls (total cholesterol: 125 +/- 12 versus 78 +/- 13 mg/dl, p = 0.0001; HDL-C 90 +/- 7 versus 55 +/- 11 mg/dl, p = 0.0001). The molar ratio of HDL-C/apoA-I was significantly lower in transgenic animals, 17 +/- 1 versus 25 +/- 2 (p = 0.0001), suggesting the increase was in smaller HDL particles. This was confirmed by native gradient gel electrophoresis. This was not due to aberrant metabolism of human apoA-I in the mouse, since human apoA-I was distributed throughout the HDL particle size range and was catabolized at the same rate as mouse apoA-I. In another study of 23 transgenic mice, HDL-C and human apoA-I levels were highly correlated (r = 0.87, p less than 0.001). The slope of the correlation line also indicated the additional HDL particles were in the smaller size range. We conclude that human apoA-I can be incorporated into mouse HDL, and excessive amounts increase HDL-C levels primarily by increasing smaller HDL particles, comparable to human HDL3 (HDL-C/apoA-I molar ratio = 18).     

Synthesis of recombinant high density lipoprotein with apolipoprotein A-I and apolipoprotein A-V

It has been shown that apolipoprotein A-V (apoA-V) over-expression significantly lowers plasma triglyceride levels and decreases atherosclerotic lesion development. To assess the feasibility of recombinant high density lipoprotein (rHDL) reconstituted with apoA-V and apolipoprotein A-I (apoA-I) as a therapeutic agent for hyperlipidemic disorder and atherosclerosis, a series of rHDL were synthesized in vitro with various mass ratios of recombinant apoA-I and apoA-V. It is interesting to find that apoA-V of rHDL had no effect on lipoprotein lipase (LPL) activation in vitro and very low density lipoprotein (VLDL) clearance in HepG2 cells and in vivo. By contrast, LPL activation and VLDL clearance were inhibited by the addition of apoA-V to rHDL. Furthermore, the apoA-V of rHDL could not redistribute from rHDL to VLDL after incubation at 37°C for 30 min. These findings suggest that an increase of apoA-V in rHDL could not play a role in VLDL clearance in vitro and in vivo, which could, at least in part, attribute to the lost redistribution of apoA-V from rHDL to VLDL and LPL binding ability of apoA-V in rHDL. The therapeutic application of rHDL reconstituted with apoA-V and apoA-I might need the construction of rHDL from which apoA-V could freely redistribute to VLDL.     

Variation at the apolipoprotein (apo) AI-CIII-AIV gene cluster and apo B gene loci is associated with lipoprotein and apolipoprotein levels in Italian children.

We have used RFLPs of the apolipoprotein (apo) B gene and apo AI-CIII-AIV gene cluster to estimate the genetic contribution of variation at these loci to the variability of plasmid lipid, lipoprotein, and apolipoprotein levels in 209 children from Sezze in central Italy. The sample was randomly divided into group I (107 children) and group II (102 children). Four site polymorphisms (PvuII, XbaI, MspI, and EcoRI) of the apo B gene and five site polymorphisms (XmnI, PstI, SstI, PvuII-CIII, and PvuII-AIV) of the apo AI-CIII-AIV gene cluster were examined in group I children. After adjustment for gender, age, and body-mass index, polymorphisms at both gene loci (PvuII-B, PvuII-CIII, and PvuII-AIV) were associated with significant effects on the levels of plasma apo AI, apo B, or high-density lipoprotein-cholesterol. RFLPs that showed significant effects in group I were genotyped in group II. All three polymorphisms were associated with similar effects on apolipoprotein levels, though for all RFLPs the magnitude of the effects was smaller in the group II children and only statistically significant for the effect of the PvuII-B genotype on apo AI levels. In the total sample of 209 children 7.4% of the sample variance in apo AI levels was explained by variation associated with the apo B PvuII-B RFLP. In addition, the PvuII-B RFLP was associated with significant effects on plasma apo B levels and explained 5.7% of the sample variance. The PvuII-CIII and PvuII-AIV polymorphisms were both associated with differences in apo AI levels, explaining 3.7%-5.7% of the sample variance. Taken together, the three PvuII polymorphisms explained 17.7% of the phenotypic variance in apo AI levels. There was significant evidence for an effect of nonlinearity of the PvuII-CIII genotypes on apo AI levels, with the individuals heterozygous for the polymorphism having the highest apo AI levels. No evidence of interaction between genotype and gender, age, and body-mass index was shown by covariance analysis. The molecular explanation of this effect is unclear. Our data show that variation at both the apo AI-CIII-AIV and apo B loci are associated with lipoprotein and apolipoprotein levels in this sample of Italian children.     

Interactions of low density lipoprotein2 and other apolipoprotein B-containing lipoproteins with lipoprotein(a)

Studies were undertaken to investigate potential interactions among plasma lipoproteins. Techniques used were low density lipoprotein2 (LDL2)-ligand blotting of plasma lipoproteins separated by nondenaturing 2.5-15% gradient gel electrophoresis, ligand binding of plasma lipoproteins by affinity chromatography with either LDL2 or lipoprotein(a) (Lp(a)) as ligands, and agarose lipoprotein electrophoresis. Ligand blotting showed that LDL2 can bind to Lp(a). When apolipoprotein(a) was removed from Lp(a) by reduction and ultracentrifugation, no interaction between LDL2 and reduced Lp(a) was detected by ligand blotting. Ligand binding showed that LDL2-Sepharose 4B columns bound plasma lipoproteins containing apolipoproteins(a), B, and other apolipoproteins. The Lp(a)-Sepharose column bound lipoproteins containing apolipoprotein B and other apolipoproteins. Furthermore, the Lp(a) ligand column bound more lipoprotein lipid than the LDL2 ligand column, with the Lp(a) ligand column having a greater affinity for triglyceride-rich lipoproteins. Lipoprotein electrophoresis of a mixture of LDL2 and Lp(a) demonstrated a single band with a mobility intermediate between that of LDL2 and Lp(a). Chemical modification of the lysine residues of apolipoprotein B (apoB) by either acetylation or acetoacetylation prevented or diminished the interaction of LDL2 with Lp(a), as shown by both agarose electrophoresis and ligand blotting using modified LDL2. Moreover, removal of the acetoacetyl group from the lysine residues of apoB by hydroxylamine reestablished the interaction of LDL2 with Lp(a). On the other hand, blocking of--SH groups of apoB by iodoacetamide failed to show any effect on the interaction between LDL2 and Lp(a). Based on these observations, it was concluded that Lp(a) interacts with LDL2 and other apoB-containing lipoproteins which are enriched in triglyceride; this interaction is due to the presence of apolipoprotein(a) and involves lysine residues of apoB interacting with the plasminogen-like domains (kringle 4) of apolipoprotein(a). Such results suggest that Lp(a) may be involved in triglyceride-rich lipoprotein metabolism, could form transient associations with apoB-containing lipoproteins in the vascular compartment, and alter the intake by the high affinity apoB, E receptor pathway.     

The linkage with apolipoprotein (a) in lipoprotein (a) modifies the immunochemical and functional properties of apolipoprotein B

Lipoprotein (a) [Lp(a)] was isolated from several donors and its apolipoprotein (a) [apo(a)] dissociated by a reductive treatment, generating the apo(a)-free form of Lp(a) [Lp(a--)] that contains apolipoprotein B (apo B) as its sole protein. Using anti-apo B monoclonal antibodies, the properties of apo B in Lp(a), Lp(a--), and autologous low-density lipoprotein (LDL) were compared. Marked differences in apo B immunoreactivity were found between these lipoproteins, due to the presence of apo(a) in Lp(a). Apo(a) enhanced the expression of two epitopes in the amino-terminal part of apo B while it diminished the immunoreactivity of three other epitopes in the LDL receptor binding domain. Accordingly, the binding of the lipoproteins to the LDL receptor was also decreased in the presence of apo(a). In a different experimental system, the incubation of antibodies that react with 27 distinct epitopes distributed along the whole length of apo B sequence with plastic-bound Lp(a) and Lp(a--) failed to reveal any epitope of apo B that is sterically hindered by the presence of apo(a). Our results demonstrate that the presence of apo(a) modified the organization and function of apo B in Lp(a) particles. The data presented indicate that most likely the modification is not due to a steric hindrance but that some more profound conformational changes are involved. We suggest that the formation of the disulfide bridge between apo B and apo(a) in Lp(a) alters the system of disulfide bonds present in apo B and thereby modifies apo B structure.     

Clusterin (complement lysis inhibitor) forms a high density lipoprotein complex with apolipoprotein A-I in human plasma

Clusterin/human complement lysis inhibitor (CLI) is incorporated stoichiometrically into the soluble terminal complement complex and inhibits the cytolytic reaction of purified complement components C5b-9 in vitro. Using an anti-clusterin affinity column, we found that an additional protein component with a molecular mass of 28-kDa co-purifies with clusterin from human plasma. We show by immunoblotting and amino acid sequencing that this component is apolipoprotein A-I (apoA-I). By using physiological salt buffers containing 0.5% Triton X-100, apoA-I is completely dissociated from clusterin bound to the antibody column. Free clusterin immobilized on the antibody-Sepharose selectively retains apoA-I from total human plasma. Delipidated apoA-I and to a lesser extent ultracentrifugation-purified high density lipoproteins (HDL) adsorbed to nitrocellulose also have a binding affinity for purified clusterin devoid of apoA-I. The isolated apoA-I-clusterin complex contains approximately 22% (w/w) lipids which are composed of 54% (mole/mol) total cholesterol (molar ratio of unesterified/esterified cholesterol, 0.58), 42% phospholipids, and 4% triglycerides. In agreement with the low lipid content, apoA-I-clusterin complexes are detected only in trace amounts in HDL fractions prepared by density ultracentrifugation. In free flow isotachophoresis, the purified apoA-I-clusterin complex has the same mobility as the native clusterin complex in human plasma and is found in the slow-migrating HDL fraction of fasting plasma. Our data indicate that clusterin circulates in plasma as a HDL complex, which may serve not only as an inhibitor of the lytic terminal complement cascade, but also as a regulator of lipid transport and local lipid redistribution.     

Physiologic mechanisms for reduced apolipoprotein A-I concentrations associated with low levels of high density lipoprotein cholesterol in patients with normal plasma lipids.

Low plasma concentrations of high density lipoprotein (HDL) cholesterol and apolipoprotein A-I (apoA-I) are major risk factors for coronary heart disease (CHD). Low HDL levels are common in patients with hypertriglyceridemia, but they also occur in those with normal plasma lipids; the latter include obese patients and cigarette smokers, though other patients with low HDL levels are neither obese nor smokers. The present study was designed to define metabolic causes of low apoA-I levels in normal-weight, normolipidemic patients. ApoA-I tracer studies were carried out in two groups of normolipidemic patients having low HDL levels to determine input rates and residence times for ApoA-I; these patients included 11 nonobese nonsmokers and 11 nonobese cigarette smokers. Their results were compared to those of 20 normal-weight, normolipidemic controls with normal HDL levels and 12 obese nonsmokers also having low HDL. In all three groups manifesting low HDL-cholesterol and low apoA-I levels, residence times for plasma apoA-I were reduced by approximately 30%, compared to control subjects with normal HDL levels. In contrast, average input rates for apoA-I were similar among the three low-HDL patients and control subjects. No differences in apoA-I kinetics were observed among any of the three groups with low apoA-I concentrations. Within each of the four groups of the study, however, input rates for apoA-I were highly correlated with plasma concentrations of apoA-I. Thus, for individuals with normal levels of plasma lipids, both residence times and input rates for apoA-I appeared to be important determinants of apoA-I levels. Residence times for apoA-I were reduced in almost all patients with low apoA-I levels, regardless of concomitant factors, whereas input rates were highly variable among individuals.     

Foxa2 activity increases plasma high density lipoprotein levels by regulating apolipoprotein M

Obesity, diabetes, insulin resistance, and hyperinsulinemia are frequently associated with a cluster of closely related lipid abnormalities such as low plasma levels of high density lipoprotein (HDL) and elevated levels of triglyceride, both known to increase the risk of developing atherosclerotic disease. The molecular mechanisms linking obesity, insulin resistance, and hyperinsulinemia to low HDL levels are incompletely understood. Here we demonstrate that insulin, through a Foxa2-mediated mechanism, inhibited the expression of apolipoprotein M (apoM), an important determinant of plasma pre-beta-HDL and alpha-HDL concentrations. Obese mice had decreased apoM expression and plasma pre-beta-HDL levels due to inactivation of Foxa2 in hyperinsulinemic states. Nuclear reexpression of Foxa2 with a phosphorylation-deficient mutant Foxa2T156A (Ad-T156A) activated apoM expression and increased plasma pre-beta-HDL and alpha-HDL levels. In contrast, haploinsufficient Foxa2(+/-) mice exhibited decreased hepatic apoM expression and plasma pre-beta-HDL and HDL levels. The increase in plasma HDL levels and pre-beta-HDL formation by Foxa2 was mediated exclusively by apoM, as constitutive active expression of Foxa2 in apoM(-/-) mice had no effect on plasma HDL levels. Our results identify a fundamental mechanism by which insulin regulates plasma HDL levels in physiological and insulin-resistant states and thus have important implications for novel therapeutic approaches to prevent atherosclerosis.     

Characterization of a mutant form of human apolipoprotein B (Thr26_Tyr27del) associated with familial hypobetalipoproteinemia

We have previously identified a deletion mutant of human apoB [apoB (Thr26_Tyr27del)] in a subject with primary hypobetalipoproteinemia. The present study determined the effect of Thr26_Tyr27del mutation on apoB secretion using transfected McA-RH7777 cells. Transient or stable transfection of apoB-48 containing the Thr26_Tyr27del mutation showed drastically reduced secretion of the mutant as compared to wild-type apoB-48. No lipoproteins containing the mutant apoB-48 were secreted into the medium. Incubation of transfected cells in a lipid-rich medium in the presence of cycloheximide showed rapid turnover of cell-associated mutant apoB-48 as compared to that of wild-type apoB-48. Immunofluorescence experiments showed that the mutant apoB-48 was mostly localized in the endoplasmic reticulum. Treatment with the proteasomal inhibitor MG132 markedly attenuated the turnover of cell-associated mutant apoB-48, whereas treatment with inhibitors of autophagosomal/lysosomal function (e.g. 3-MA or ammonium chloride) had no effect. Taken together, these results indicated that the defective secretion of the Thr26_Tyr27del mutant was associated with increased intracellular degradation of apoB through the proteasome-dependent pathway.     

Isolation and partial characterization of apolipoprotein (a) from human lipoprotein (a)

A procedure was developed for the dissociation of apolipoprotein (a) (apo (a)) from pure human lipoprotein (a) (Lp(a)) prepared by density gradient ultracentrifugation and gel filtration. Lp(a) was ultracentrifuged through a layer of saline which was adjusted to a density of 1.182 g/mL and contained 30 mM dithiothreitol (50 mM) and phenylmethylsulfonyl fluoride (1.25 mM). Following centrifugation, the lipid and apolipoprotein B (apo B) were recovered as a lipoprotein (Lp(a) B) in the supernatant fraction, while the apo (a) was recovered as a lipid-poor protein pellet. An investigation of the supernatant lipoprotein by electron microscopy and compositional analysis revealed that it was similar in size and composition to low density lipoprotein (LDL) isolated from the same density range and contained apo B100 with an amino acid and carbohydrate composition which was similar to apo B from LDL. Estimates of the apparent molecular weight of the apo (a) varied amongst individuals but was always greater than apo B100 (congruent to 450,000). The amino acid composition of apo (a), which was very distinct from apo B, was characterized by a higher content of serine, threonine, proline, and tyrosine, but lower amounts of isoleucine, phenylalanine, and lysine when compared with apo B of Lp(a) or LDL. The apo (a) contained a much higher proportion of carbohydrate, in particular N-acetylgalactosamine, galactose, and N-acetylneuraminic acid (which were three- to six-fold higher) than the apo B of Lp(a). It is concluded that apo (a) is distinct from other apolipoproteins owing to its low avidity for lipid and the nature of the interaction with apo B. Lp(a) consists of an LDL-like particle with a carbohydrate-rich apo (a) attached to the surface of apo B.     

Assembly of lipoprotein (a) in transgenic rabbits expressing human apolipoprotein (a)

The study of human lipoprotein (a) [Lp(a)] has been hampered due to the lack of appropriate animal models since apolipoprotein (a) [apo(a)] is found only in primates and humans. In addition, human apo(a) in transgenic mice can not bind to murine apoB to form Lp(a) particles. In this study, we generated three independent transgenic rabbits expressing human apo(a) in their plasma at 1.8-4.5 mg/dl. In the plasma of transgenic rabbits, unlike the plasma of transgenic mice, about 80% of the apo(a) was covalently associated with rabbit apo-B and was contained in the fractions with density 1.02-1.10 g/ml, indicating the formation of Lp(a). These results suggest that transgenic rabbits expressing human apo(a) exhibit efficient assembly of Lp(a) and can be used as an animal model for the study of human Lp(a).     

Impaired platelet activation in familial high density lipoprotein deficiency (Tangier disease)

ATP binding cassette transporter A1 (ABCA1) is involved in regulation of intracellular lipid trafficking and export of cholesterol from cells to high density lipoproteins. ABCA1 defects cause Tangier disease, a disorder characterized by absence of high density lipoprotein and thrombocytopenia. In the present study we have demonstrated that ABCA1 is expressed in human platelets and that fibrinogen binding and CD62 surface expression in response to collagen and low concentrations of thrombin, but not to ADP, are defective in platelets from Tangier patients and ABCA1-deficient animals. The expression of platelet membrane receptors such as GPVI, alpha2beta1 integrin, and GPIIb/IIIa, the collagen-induced changes in phosphatidylserine and cholesterol distribution, and the collagen-induced signal transduction examined by phosphorylation of LAT and p72syk and by intracellular Ca2+ mobilization were unaltered in Tangier platelets. The electron microscopy of Tangier platelets revealed reduced numbers of dense bodies and the presence of giant granules typically encountered in platelets from Chediak-Higashi syndrome. Further studies demonstrated impaired release of dense body content in platelets from Tangier patients and ABCA1-deficient animals. In addition, Tangier platelets were characterized by defective surface exposure of dense body and lysosomal markers (CD63, LAMP-1, LAMP-2, CD68) during collagen- and thrombin-induced stimulation and by abnormally high lysosomal pH. We conclude that intact ABCA1 function is necessary for proper maturation of dense bodies in platelets. The impaired release of the content of dense bodies may explain the defective activation of Tangier platelets by collagen and low concentrations of thrombin, but not by ADP.